Social media platforms, particularly TikTok, have become an integral part of daily life for millions globally, generating vast amounts of user-generated content. For businesses and content creators, understanding the sentiments behind these user comments is crucial for gauging public perception and refining their commercial or marketing strategies. However, the unique nature of comments on TikTok, characterized by their brevity, informal language, slang, and emojis, presents significant challenges for sentiment analysis. We developed a sentiment analysis system to address these challenges using advanced Natural Language Processing (NLP) and Deep Learning (DL) techniques. The system’s architecture, combining a frontend built with Angular and Tailwind CSS and a backend powered by FastAPI and a fine-tuned RoBERTa-based model, allows for real-time analysis of large datasets. Our model, trained with PyTorch CUDA, achieved a high accuracy of 87.2%, with a precision of 90%, a recall of 83.6%, and an F1-score of 86.7%.